Data Science Foundation - Assignment on Facebook Analysis

Analysis Attempted By : Mohit Dewan

Domain: Facebook - Social media analysis on usage.

Project Overview : Facebook is a popular free social networking website that allows registered users to create profiles, upload photos and video, send messages and keep in touch with friends, family and colleagues. A lot of data are generated on Facebook daily. Emphasis is on looking at the various trends e.g. friend count, tenure, date of brith and distribution of the parameters in our data.

---

Table of Contents

---

1. Introduction
2. Problem Statement
3. Installing & Importing Libraries

• 3.1 Installing Libraries
• 3.2 Upgrading Libraries
• 3.3 Importing Libraries

4. Data Acquisition & Description

• 4.1 Data Description
• 4.2 Data Information

5. Data Pre-profiling
6. Data Pre-Processing
7. Data Post-profiling
8. Exploratory Data Analysis
9. Summarization

---

1. Introduction

---

• Facebook is an online social media and social networking serviceowned by American technology giant Meta Platforms.

• Facebook is a social media platform that was founded in 2004 by Mark Zuckerberg, Eduardo Saverin, Andrew McCollum, Dustin Moskovitz, and Chris Hughes.

• Initially designed for college students, it quickly grew in popularity and is now used by people all over the world.

• As of 2020, Facebook's has the **second most valuable media brand worldwide**.

  

---

2. Problem Statement

---

• Facebook wants to expand their business globally in the near future.

• As of December 2022, Facebook claimed **2.96 billion** monthly active users and ranked third worldwide among the most visited websites.

• Facebook hired a team of data scientist to analyize the dataset provided and to identify certain patterns with respect to how the users are making use of this most popular social networking app depending on their age group,gender etc..

• Based on Data Analysis outcome what can be showcased for their business from the dataset using EDA.

---

3. Installing & Importing Libraries

---

---

3.1 Installing Libraries

---

• This section is emphasised on installing the necessary libraries that will be required for Analysis.

!pip install -q datascience                                         
!pip install -q pandas-profiling                                    
!pip install -q yellowbrick

---

3.2 Upgrading Libraries

---

• This section is emphasised on upgrading the necessary libraries that will be required for Analysis.

!pip install -q --upgrade pandas-profiling
!pip install -q --upgrade yellowbrick
!pip install --upgrade scikit-learn

Defaulting to user installation because normal site-packages is not writeable
Requirement already satisfied: scikit-learn in c:\programdata\anaconda3\lib\site-packages (1.2.1)
Collecting scikit-learn
  Downloading scikit_learn-1.2.2-cp310-cp310-win_amd64.whl (8.3 MB)
     ---------------------------------------- 8.3/8.3 MB 20.4 MB/s eta 0:00:00
Requirement already satisfied: numpy>=1.17.3 in c:\programdata\anaconda3\lib\site-packages (from scikit-learn) (1.23.5)
Requirement already satisfied: scipy>=1.3.2 in c:\users\pankaj dewan\appdata\roaming\python\python310\site-packages (from scikit-learn) (1.9.3)
Requirement already satisfied: threadpoolctl>=2.0.0 in c:\programdata\anaconda3\lib\site-packages (from scikit-learn) (2.2.0)
Requirement already satisfied: joblib>=1.1.1 in c:\programdata\anaconda3\lib\site-packages (from scikit-learn) (1.1.1)
Installing collected packages: scikit-learn
Successfully installed scikit-learn-1.2.2

---

3.3 Importing Libraries

---

• This section is emphasised on installing and importing the necessary libraries that will be required for Analysis.

import numpy as np                                                  # For numerical python operations
#-------------------------------------------------------------------------------------------------------------------------------
import pandas as pd                                                 # Importing for panel data analysis
#-------------------------------------------------------------------------------------------------------------------------------
import matplotlib.pyplot as plt                                     # Importing pyplot interface using matplotlib
import seaborn as sns                                               # Importing seaborm library for interactive visualization
%matplotlib inline
#-------------------------------------------------------------------------------------------------------------------------------
import warnings                                                     # Importing warning to disable runtime warnings
warnings.filterwarnings("ignore")                                   # Warnings will appear only once

---

4. Data Acquisition & Wrangling

---

• This section is lays emphasis on the accquiring the data and obtain some descriptive information out of it.
• The data is locally available.
• This dataset provides a usage analysis of various facebook users such as the tenure of their account,number of friends they have,number of friendships they have initiated , the number of likes received by them through the FB website and through the mobile app. -The dataset comprises of 99003 observations of 15 columns. Below is a table showing names of all the columns and their description.
• Before moving ahead lets have an overview about the features of the data set :

Id	Features	Description
01	UserID	A numeric value uniquely identifying the user.
02	Age	Age of the user in years.
03	DOB_Day	Day part of the user's date of birth.
04	DOB_Year	Year part of the user's date of birth.
05	DOB_Month	Month part of the user's date of birth.
06	Gender	Gender of the user.
07	Tenure	Number of days since the user has been on FB.
08	Friend Count	Number of friends the user has.
09	friendships_initiated	Number of friendships initiated by the user.
10	likes	Total number of posts liked by the user.
11	likes received	Total Number of likes received by user's posts.
12	mobile likes	Number of posts liked by the user through mobile app.
13	mobile likes received	Number of likes received by user through mobile app.
14	www likes	Number of posts liked by the user through web.
15	www likes received	Number of likes received by user through web.

# Importing and Reading the data
fb_df = pd.read_csv(r"E:\fb_data.csv")
fb_df
print('Data Shape: ', fb_df.shape)

Data Shape:  (99003, 15)

fb_df.head(5)

	userid	age	dob_day	dob_year	dob_month	gender	tenure	friend_count	friendships_initiated	likes	likes_received	mobile_likes	mobile_likes_received	www_likes	www_likes_received
0	2094382	14	19	1999	11	male	266.0	0	0	0	0	0	0	0	0
1	1192601	14	2	1999	11	female	6.0	0	0	0	0	0	0	0	0
2	2083884	14	16	1999	11	male	13.0	0	0	0	0	0	0	0	0
3	1203168	14	25	1999	12	female	93.0	0	0	0	0	0	0	0	0
4	1733186	14	4	1999	12	male	82.0	0	0	0	0	0	0	0	0

4.1 Data Description

• In this section we will get information about the data and see some observations.
• To get some quick description out of the data you can use describe method defined in pandas library

fb_df.describe()

	userid	age	dob_day	dob_year	dob_month	tenure	friend_count	friendships_initiated	likes	likes_received	mobile_likes	mobile_likes_received	www_likes	www_likes_received
count	9.900300e+04	99003.000000	99003.000000	99003.000000	99003.000000	99001.000000	99003.000000	99003.000000	99003.000000	99003.000000	99003.000000	99003.000000	99003.000000	99003.000000
mean	1.597045e+06	37.280224	14.530408	1975.719776	6.283365	537.887375	196.350787	107.452471	156.078785	142.689363	106.116300	84.120491	49.962425	58.568831
std	3.440592e+05	22.589748	9.015606	22.589748	3.529672	457.649874	387.304229	188.786951	572.280681	1387.919613	445.252985	839.889444	285.560152	601.416348
min	1.000008e+06	13.000000	1.000000	1900.000000	1.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
25%	1.298806e+06	20.000000	7.000000	1963.000000	3.000000	226.000000	31.000000	17.000000	1.000000	1.000000	0.000000	0.000000	0.000000	0.000000
50%	1.596148e+06	28.000000	14.000000	1985.000000	6.000000	412.000000	82.000000	46.000000	11.000000	8.000000	4.000000	4.000000	0.000000	2.000000
75%	1.895744e+06	50.000000	22.000000	1993.000000	9.000000	675.000000	206.000000	117.000000	81.000000	59.000000	46.000000	33.000000	7.000000	20.000000
max	2.193542e+06	113.000000	31.000000	2000.000000	12.000000	3139.000000	4923.000000	4144.000000	25111.000000	261197.000000	25111.000000	138561.000000	14865.000000	129953.000000

fb_df.describe(include='all')

	userid	age	dob_day	dob_year	dob_month	gender	tenure	friend_count	friendships_initiated	likes	likes_received	mobile_likes	mobile_likes_received	www_likes	www_likes_received
count	9.900300e+04	99003.000000	99003.000000	99003.000000	99003.000000	98828	99001.000000	99003.000000	99003.000000	99003.000000	99003.000000	99003.000000	99003.000000	99003.000000	99003.000000
unique	NaN	NaN	NaN	NaN	NaN	2	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
top	NaN	NaN	NaN	NaN	NaN	male	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
freq	NaN	NaN	NaN	NaN	NaN	58574	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
mean	1.597045e+06	37.280224	14.530408	1975.719776	6.283365	NaN	537.887375	196.350787	107.452471	156.078785	142.689363	106.116300	84.120491	49.962425	58.568831
std	3.440592e+05	22.589748	9.015606	22.589748	3.529672	NaN	457.649874	387.304229	188.786951	572.280681	1387.919613	445.252985	839.889444	285.560152	601.416348
min	1.000008e+06	13.000000	1.000000	1900.000000	1.000000	NaN	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
25%	1.298806e+06	20.000000	7.000000	1963.000000	3.000000	NaN	226.000000	31.000000	17.000000	1.000000	1.000000	0.000000	0.000000	0.000000	0.000000
50%	1.596148e+06	28.000000	14.000000	1985.000000	6.000000	NaN	412.000000	82.000000	46.000000	11.000000	8.000000	4.000000	4.000000	0.000000	2.000000
75%	1.895744e+06	50.000000	22.000000	1993.000000	9.000000	NaN	675.000000	206.000000	117.000000	81.000000	59.000000	46.000000	33.000000	7.000000	20.000000
max	2.193542e+06	113.000000	31.000000	2000.000000	12.000000	NaN	3139.000000	4923.000000	4144.000000	25111.000000	261197.000000	25111.000000	138561.000000	14865.000000	129953.000000

fb_df.nunique() #this function result the unique value of each column

userid                   99003
age                        101
dob_day                     31
dob_year                   101
dob_month                   12
gender                       2
tenure                    2426
friend_count              2562
friendships_initiated     1519
likes                     2924
likes_received            2681
mobile_likes              2396
mobile_likes_received     2004
www_likes                 1726
www_likes_received        1636
dtype: int64

fb_df.columns  #this function show the headers of all coloumn

Index(['userid', 'age', 'dob_day', 'dob_year', 'dob_month', 'gender', 'tenure',
       'friend_count', 'friendships_initiated', 'likes', 'likes_received',
       'mobile_likes', 'mobile_likes_received', 'www_likes',
       'www_likes_received'],
      dtype='object')

MISSING VALUES 1636

Sometimes it may happen that there might be some missing data or some inconsistency in the data which may create problems further while analyzing it so it is important you replace the missing values or the null values.

Normally there are two ways that can be followed

Either you delete the entire row which has missing data. You fill it with the mean, mode, or median of the data present in the column. I personally prefer the second way because deleting the row may lead to losing the data and while analyzing or while training it is good to have as much data as possible for predicting better results.

The below code shows us how many null values are present in each of the columns in the data frame.

Gender: 175 Null values

Tenure: 2 Null values

pd.isnull(fb_df["gender"]).sum()

175

fb_df.columns

Index(['userid', 'age', 'dob_day', 'dob_year', 'dob_month', 'gender', 'tenure',
       'friend_count', 'friendships_initiated', 'likes', 'likes_received',
       'mobile_likes', 'mobile_likes_received', 'www_likes',
       'www_likes_received'],
      dtype='object')

fb_df['gender'].value_counts()

male      58574
female    40254
Name: gender, dtype: int64

fb_df['gender']=fb_df['gender'].fillna(fb_df['gender'].mode()[0])

fb_df['gender'].value_counts()

male      58749
female    40254
Name: gender, dtype: int64

pd.isnull(fb_df["gender"]).sum()

0

import matplotlib.pyplot as plt
import seaborn as sns
sns.distplot(fb_df.userid)
plt.show()

fb_df.userid.quantile([0.0])[0]

1000008.0

fb_df.userid.quantile([0.25])

0.25    1298805.5
Name: userid, dtype: float64

fb_df.userid.quantile([0.25]).index

Float64Index([0.25], dtype='float64')

fb_df.userid.quantile([0.25]).values

array([1298805.5])

sns.distplot(fb_df.userid)
plt.axvline(fb_df.userid.quantile([0.0])[0], color='r')
plt.axvline(fb_df.userid.quantile([0.25])[0.25], color='r')
plt.axvline(fb_df.userid.quantile([0.50])[0.50], color='r') # median
plt.axvline(fb_df.userid.quantile([0.75])[0.75], color='r')
plt.axvline(fb_df.userid.quantile([1.0])[1.0], color='r')
plt.show()

median_value = fb_df['tenure'].median()
print (median_value)

412.0

fb_df['tenure'].fillna(value=median_value, inplace=True)

pd.isnull(fb_df["tenure"]).sum()

0

fb_df.isnull().sum()    # to check if all missing values are handled

userid                   0
age                      0
dob_day                  0
dob_year                 0
dob_month                0
gender                   0
tenure                   0
friend_count             0
friendships_initiated    0
likes                    0
likes_received           0
mobile_likes             0
mobile_likes_received    0
www_likes                0
www_likes_received       0
dtype: int64

4.2 Data Information

• In this section we will see the information about the types of features.

fb_df.info() #This function gives information about the columns type

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 99003 entries, 0 to 99002
Data columns (total 15 columns):
 #   Column                 Non-Null Count  Dtype  
---  ------                 --------------  -----  
 0   userid                 99003 non-null  int64  
 1   age                    99003 non-null  int64  
 2   dob_day                99003 non-null  int64  
 3   dob_year               99003 non-null  int64  
 4   dob_month              99003 non-null  int64  
 5   gender                 99003 non-null  object 
 6   tenure                 99003 non-null  float64
 7   friend_count           99003 non-null  int64  
 8   friendships_initiated  99003 non-null  int64  
 9   likes                  99003 non-null  int64  
 10  likes_received         99003 non-null  int64  
 11  mobile_likes           99003 non-null  int64  
 12  mobile_likes_received  99003 non-null  int64  
 13  www_likes              99003 non-null  int64  
 14  www_likes_received     99003 non-null  int64  
dtypes: float64(1), int64(13), object(1)
memory usage: 11.3+ MB

  info function gives us the following insights into the dataframe:
	There are a total of 99003 samples (rows) and 15 columns in the dataframe.
	There are 13 columns with a numeric datatype.
	There is a float and Category column each.
	There are missing values in the data.

fb_df.isna().sum().sort_values(ascending=False)

userid                   0
age                      0
dob_day                  0
dob_year                 0
dob_month                0
gender                   0
tenure                   0
friend_count             0
friendships_initiated    0
likes                    0
likes_received           0
mobile_likes             0
mobile_likes_received    0
www_likes                0
www_likes_received       0
dtype: int64

fb_df.head(5)

	userid	age	dob_day	dob_year	dob_month	gender	tenure	friend_count	friendships_initiated	likes	likes_received	mobile_likes	mobile_likes_received	www_likes	www_likes_received
0	2094382	14	19	1999	11	male	266.0	0	0	0	0	0	0	0	0
1	1192601	14	2	1999	11	female	6.0	0	0	0	0	0	0	0	0
2	2083884	14	16	1999	11	male	13.0	0	0	0	0	0	0	0	0
3	1203168	14	25	1999	12	female	93.0	0	0	0	0	0	0	0	0
4	1733186	14	4	1999	12	male	82.0	0	0	0	0	0	0	0	0

fb_df.gender.unique()

array(['male', 'female'], dtype=object)

# To check which column is unique, null and not-null values. If unique found then set as Index column
fb_df.agg(['count', 'size', 'nunique'])

	userid	age	dob_day	dob_year	dob_month	gender	tenure	friend_count	friendships_initiated	likes	likes_received	mobile_likes	mobile_likes_received	www_likes	www_likes_received
count	99003	99003	99003	99003	99003	98828	99001	99003	99003	99003	99003	99003	99003	99003	99003
size	99003	99003	99003	99003	99003	99003	99003	99003	99003	99003	99003	99003	99003	99003	99003
nunique	99003	101	31	101	12	2	2426	2562	1519	2924	2681	2396	2004	1726	1636

## Data Structure
## The data frame has 99003 observations and 15 variables. 
## Apart from "gender" all the other variables are integers and tenure as float as 2 records are null.

---

5. Data Pre-Profiling

---

• To gain insights from data we must look into each aspect of it very carefully. We will start with observing few rows and columns of data both from the starting and from the end.

import pandas as pd
from ydata_profiling import ProfileReport

profile = fb_df.profile_report(title = 'Pre Profile Facebook Dataset')
profile.to_file(output_file='Pre Profile Facebook Data Analysis.html')

Summarize dataset:   0%|          | 0/5 [00:00<?, ?it/s]

Generate report structure:   0%|          | 0/1 [00:00<?, ?it/s]

Render HTML:   0%|          | 0/1 [00:00<?, ?it/s]

Export report to file:   0%|          | 0/1 [00:00<?, ?it/s]

Observations from Pandas Profiling before Data Processing

Dataset info:

Number of variables:15 Number of observations: 99003 Missing cells: <0.1% Variables types:

Numeric = 14 Categorical = 1 Dataset has 175 (0.2%) Missing Value in gender column and 2 in tenure column

There are Zeros in the following Columns

friend_count friendships_initiated likes likes_received mobile_likes mobile_likes_received www_likes www_likes_received

---

6. Data Pre-Processing

---

fb_df.drop('userid',axis =1,inplace=True)

fb_df['gender'].mode()

0    male
Name: gender, dtype: object

fb_df['gender'] = fb_df['gender'].replace(np.NaN,'male')

fb_df['gender'].unique()

array(['male', 'female'], dtype=object)

fb_df['tenure'].median()

412.0

fb_df['tenure'] = fb_df['tenure'].replace(np.NaN,412.0)

fb_df.isnull().sum().sort_values(ascending = False)

age                      0
dob_day                  0
dob_year                 0
dob_month                0
gender                   0
tenure                   0
friend_count             0
friendships_initiated    0
likes                    0
likes_received           0
mobile_likes             0
mobile_likes_received    0
www_likes                0
www_likes_received       0
dtype: int64

# create date_of_birth column using these variables: dob_year , dob_month , and dob_day

# Before convereting day, month and year column to Date Of birth, need to check date validity for each column
print("Feature 'DOB_Day' has '{unique_values}' unique values".format(unique_values=np.sort(fb_df["dob_day"].unique())))
print("Feature 'DOB_Month' has '{unique_values}' unique values".format(unique_values=np.sort(fb_df["dob_month"].unique())))
print("Feature 'DOB_Year' has '{unique_values}' unique values".format(unique_values=np.sort(fb_df["dob_year"].unique())))

Feature 'DOB_Day' has '[ 1  2  3  4  5  6  7  8  9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
 25 26 27 28 29 30 31]' unique values
Feature 'DOB_Month' has '[ 1  2  3  4  5  6  7  8  9 10 11 12]' unique values
Feature 'DOB_Year' has '[1900 1901 1902 1903 1904 1905 1906 1907 1908 1909 1910 1911 1912 1913
 1914 1915 1916 1917 1918 1919 1920 1921 1922 1923 1924 1925 1926 1927
 1928 1929 1930 1931 1932 1933 1934 1935 1936 1937 1938 1939 1940 1941
 1942 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955
 1956 1957 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969
 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983
 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997
 1998 1999 2000]' unique values

# As day, month and year column data valid now convert into date type column 'DateOfBirth'
fb_df.insert(1,"DateOfBirth",pd.to_datetime(fb_df.dob_year*10000+fb_df.dob_month*100+fb_df.dob_day,format='%Y%m%d'))

fb_df.head()

	age	DateOfBirth	date_of_birth	dob_day	dob_year	dob_month	gender	tenure	friend_count	friendships_initiated	likes	likes_received	mobile_likes	mobile_likes_received	www_likes	www_likes_received
0	14	1999-11-19	1999-11-19	19	1999	11	male	266.0	0	0	0	0	0	0	0	0
1	14	1999-11-02	1999-11-02	2	1999	11	female	6.0	0	0	0	0	0	0	0	0
2	14	1999-11-16	1999-11-16	16	1999	11	male	13.0	0	0	0	0	0	0	0	0
3	14	1999-12-25	1999-12-25	25	1999	12	female	93.0	0	0	0	0	0	0	0	0
4	14	1999-12-04	1999-12-04	4	1999	12	male	82.0	0	0	0	0	0	0	0	0

fb_df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 99003 entries, 0 to 99002
Data columns (total 16 columns):
 #   Column                 Non-Null Count  Dtype         
---  ------                 --------------  -----         
 0   age                    99003 non-null  int64         
 1   DateOfBirth            99003 non-null  datetime64[ns]
 2   date_of_birth          99003 non-null  datetime64[ns]
 3   dob_day                99003 non-null  int64         
 4   dob_year               99003 non-null  int64         
 5   dob_month              99003 non-null  int64         
 6   gender                 99003 non-null  object        
 7   tenure                 99003 non-null  float64       
 8   friend_count           99003 non-null  int64         
 9   friendships_initiated  99003 non-null  int64         
 10  likes                  99003 non-null  int64         
 11  likes_received         99003 non-null  int64         
 12  mobile_likes           99003 non-null  int64         
 13  mobile_likes_received  99003 non-null  int64         
 14  www_likes              99003 non-null  int64         
 15  www_likes_received     99003 non-null  int64         
dtypes: datetime64[ns](2), float64(1), int64(12), object(1)
memory usage: 12.1+ MB

fb_df.describe(include = 'all')

	age	DateOfBirth	date_of_birth	dob_day	dob_year	dob_month	gender	tenure	friend_count	friendships_initiated	likes	likes_received	mobile_likes	mobile_likes_received	www_likes	www_likes_received
count	99003.000000	99003	99003	99003.000000	99003.000000	99003.000000	99003	99003.000000	99003.000000	99003.000000	99003.000000	99003.000000	99003.000000	99003.000000	99003.000000	99003.000000
unique	NaN	23151	23151	NaN	NaN	NaN	2	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
top	NaN	1988-01-01 00:00:00	1988-01-01 00:00:00	NaN	NaN	NaN	male	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
freq	NaN	656	656	NaN	NaN	NaN	58749	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
first	NaN	1900-01-01 00:00:00	1900-01-01 00:00:00	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
last	NaN	2000-10-27 00:00:00	2000-10-27 00:00:00	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
mean	37.280224	NaN	NaN	14.530408	1975.719776	6.283365	NaN	537.884832	196.350787	107.452471	156.078785	142.689363	106.116300	84.120491	49.962425	58.568831
std	22.589748	NaN	NaN	9.015606	22.589748	3.529672	NaN	457.645601	387.304229	188.786951	572.280681	1387.919613	445.252985	839.889444	285.560152	601.416348
min	13.000000	NaN	NaN	1.000000	1900.000000	1.000000	NaN	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
25%	20.000000	NaN	NaN	7.000000	1963.000000	3.000000	NaN	226.000000	31.000000	17.000000	1.000000	1.000000	0.000000	0.000000	0.000000	0.000000
50%	28.000000	NaN	NaN	14.000000	1985.000000	6.000000	NaN	412.000000	82.000000	46.000000	11.000000	8.000000	4.000000	4.000000	0.000000	2.000000
75%	50.000000	NaN	NaN	22.000000	1993.000000	9.000000	NaN	675.000000	206.000000	117.000000	81.000000	59.000000	46.000000	33.000000	7.000000	20.000000
max	113.000000	NaN	NaN	31.000000	2000.000000	12.000000	NaN	3139.000000	4923.000000	4144.000000	25111.000000	261197.000000	25111.000000	138561.000000	14865.000000	129953.000000

---

7. Data Post-Profiling

---

• To gain insights from post processed data we must look into each aspect of it very carefully. We will start with observing few rows and columns of data both from the starting and from the end.

profile = fb_df.profile_report(title = 'Post Profile Facebook Data')
profile.to_file(output_file='Post Profile Facebook Data Analysis after Processing.html')

Summarize dataset:   0%|          | 0/5 [00:00<?, ?it/s]

Generate report structure:   0%|          | 0/1 [00:00<?, ?it/s]

Render HTML:   0%|          | 0/1 [00:00<?, ?it/s]

Export report to file:   0%|          | 0/1 [00:00<?, ?it/s]

Now we have preprocessed the data, now the dataset doesnot contain missing values, we have also introduced new feature named date_of_birth. So, the pandas profiling report which we have generated after preprocessing will give us more beneficial insights. You can compare the two reports, i.e Post Profile Facebook Data Analysis after Processing.html and Pre Profile Facebook Data Analysis before Processing.html.<br/

observations:

• In the Dataset info, Total Missing(%) = 0.0%
• Number of variables = 15
• Observe the newly created variable date_of_birth.
• Dataset has 8 (<0.1%) duplicate rows, needs to be removed from the datatset

Data Post-Processing.

fb_df.drop_duplicates(inplace=True)

fb_df.describe(include='all')

	age	DateOfBirth	date_of_birth	dob_day	dob_year	dob_month	gender	tenure	friend_count	friendships_initiated	likes	likes_received	mobile_likes	mobile_likes_received	www_likes	www_likes_received
count	98995.000000	98995	98995	98995.000000	98995.000000	98995.000000	98995	98995.000000	98995.000000	98995.000000	98995.000000	98995.000000	98995.000000	98995.000000	98995.000000	98995.000000
unique	NaN	23151	23151	NaN	NaN	NaN	2	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
top	NaN	1988-01-01 00:00:00	1988-01-01 00:00:00	NaN	NaN	NaN	male	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
freq	NaN	651	651	NaN	NaN	NaN	58741	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
first	NaN	1900-01-01 00:00:00	1900-01-01 00:00:00	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
last	NaN	2000-10-27 00:00:00	2000-10-27 00:00:00	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN	NaN
mean	37.281146	NaN	NaN	14.531502	1975.718854	6.283792	NaN	537.916986	196.366625	107.461124	156.091399	142.700894	106.124875	84.127289	49.966463	58.573564
std	22.590414	NaN	NaN	9.015150	22.590414	3.529495	NaN	457.646219	387.315871	188.792125	572.302084	1387.975100	445.269954	839.923040	285.571337	601.440418
min	13.000000	NaN	NaN	1.000000	1900.000000	1.000000	NaN	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000	0.000000
25%	20.000000	NaN	NaN	7.000000	1963.000000	3.000000	NaN	226.000000	31.000000	17.000000	1.000000	1.000000	0.000000	0.000000	0.000000	0.000000
50%	28.000000	NaN	NaN	14.000000	1985.000000	6.000000	NaN	412.000000	82.000000	46.000000	11.000000	8.000000	4.000000	4.000000	0.000000	2.000000
75%	50.000000	NaN	NaN	22.000000	1993.000000	9.000000	NaN	675.000000	206.000000	117.000000	81.000000	59.000000	46.000000	33.000000	7.000000	20.000000
max	113.000000	NaN	NaN	31.000000	2000.000000	12.000000	NaN	3139.000000	4923.000000	4144.000000	25111.000000	261197.000000	25111.000000	138561.000000	14865.000000	129953.000000

8. Exploratory Data Analysis

---

How the Age of Users Distributed

labels = ['12-14', '15-20','21-30','31-40','41-50', '51-60', '61-70', '71-80','81-90','91-100','101-110','111-120']
fb_df['age_group'] = pd.cut(fb_df['age'],
                         [10,15,20,30,40,50,60,70,80,90,100,110,120],
                         labels= labels, include_lowest=True)

sns.set_style('whitegrid')
plt.figure(figsize=(10,5))
sns.countplot(x='age_group',data=fb_df)

<AxesSubplot: xlabel='age_group', ylabel='count'>

sns.set_style('whitegrid')
plt.figure(figsize=(20,12))
plt.xticks(rotation=45)
sns.countplot(x='age',data=fb_df)

<AxesSubplot: xlabel='age', ylabel='count'>

• From above graph it shows 103 ,108 and 113 age group, user count is higher compare to other years, might be user is providing incorrect detail by not specifying their year.

fb_df.groupby(['age_group'])['age_group'].count()

age_group
12-14       5027
15-20      19725
21-30      28639
31-40      12490
41-50       8968
51-60       9319
61-70       6855
71-80       2249
81-90        817
91-100      1219
101-110     3449
111-120      238
Name: age_group, dtype: int64

• Maximun users of Facebook are of Age 15 to 30 years.
• Maximum Age:113 years and Minimum Age:13 years

Gender wise Analysis

fb_df['gender'].value_counts().plot(kind='pie',explode=[0.02,0.02],fontsize=14, autopct='%1.1f%%', 
                                               figsize=(8,8), shadow=True, startangle=135, legend=True, cmap='summer')
plt.title('Pie chart showing the Gender wise Facebook Users Status')

Text(0.5, 1.0, 'Pie chart showing the Gender wise Facebook Users Status')

• Maximum users of Facebook are Males about 59.3%.

Analysis based on Date of Birth

We have converted Date Of birth into small age group of 10 to cover entire age population.

Using the age groups bins we will perform analysis on user count, genderwise user count, tenure, Friend_count etc.

# The following analysis we are doing against Date Of birth and Gender
labels = ['11-20', '21-30', '31-40', '41-50', '51-60', '61-70', '71-80','81-90','91-100','101-110','111-120']
fb_df['age_group'] = pd.cut(fb_df.age, [10,20,30,40,50,60,70,80,90,100,110,120], right=True, labels=labels)
fb_df['age_group'] = fb_df['age_group'].astype('category')
facebook = fb_df.copy()

sns.set()
groupByAgeGroupAndGender = fb_df.groupby(['age_group','gender']).size().reset_index(name='counts')
with sns.axes_style('white'):
    fb_df = sns.catplot(x="age_group", y=None ,hue ='gender',data=fb_df, kind="count",height=8, color='blue', aspect=2)
    sns.pointplot(x='age_group', y='counts', data=groupByAgeGroupAndGender, color='r')

plt.title("Distribution of genderwise age group against user count")
plt.ylabel('count of facebook users')
plt.xlabel('age group');
plt.show()

It shows between age of 11 to 30 years users are using facebook twice more than 31 ownwards age group

It shows after age of 31 years many people are not using facebook as networking media. But age of 101 to 110 users are using more compare to other age group between 81-to-100. May be user has provided default date 1900 year while setting up the facebook profile so that age can be unknown/ignored.

Facebook used by Male is higher than Female user between 11 to 50 year age but after 51 age ownards female usage is gradually increased in comparison to male population.

facebook = facebook.reset_index()

var = 'age'
fig, axes =plt.subplots(4,3, figsize=(15,12))
axes = axes.flatten()
for ax, i in zip(axes,labels):
    sns.countplot(y = var, data = facebook[facebook["age_group"] == i] ,ax=ax)
plt.tight_layout()  
plt.show()

From above graph it shows every 3rd and some age-group 8th year user count is higher compare to other years. Either users are providing incorrect detail by not specifying their year.

Analysis based on Tenure

sns.set(color_codes=True)
plt.figure(figsize=(20,12))
sns.set_palette(sns.color_palette("Set1", n_colors=5, desat=.5))
sns.distplot(fb_df["tenure"])

<AxesSubplot: xlabel='tenure', ylabel='Density'>

• The above plot shocases more number of User have joined from Last 2 Years of data collection.

Friend Count Distribution

fb_df["friend_count"].value_counts()

0       1956
1       1815
2       1116
3        860
5        789
        ... 
2601       1
2921       1
4066       1
2647       1
2002       1
Name: friend_count, Length: 2562, dtype: int64

Percentage_of_friend_count_nill = fb_df['friend_count'].value_counts().max() / (fb_df.friend_count.count())*100
print('Percentage of Zero Friend Count = ', Percentage_of_friend_count_nill.round(decimals=2))

Percentage of Zero Friend Count =  1.98

sns.set(color_codes=True)
plt.figure(figsize=(18,9))
sns.set_palette(sns.color_palette("muted"))
sns.set(font_scale=1.5)
sns.distplot(fb_df["friend_count"])

<AxesSubplot: xlabel='friend_count', ylabel='Density'>

• more number of users have fewer than 500 Facebook Friends
• 2% of users did not have any friends

Distribution of Friendships Initiated

plt.figure(figsize=(18,9))
sns.set_palette(sns.color_palette("deep"))
sns.set(font_scale=1.5)
sns.distplot(fb_df["friendships_initiated"])

<AxesSubplot: xlabel='friendships_initiated', ylabel='Density'>

fb_df['friendships_initiated'].value_counts()

0       2991
1       2211
2       1550
3       1355
4       1352
        ... 
1987       1
2012       1
2213       1
1804       1
1524       1
Name: friendships_initiated, Length: 1519, dtype: int64

Percentage_of_friendships_Initiated_nill = fb_df['friendships_initiated'].value_counts().max() / (fb_df.friendships_initiated.count())*100
print('Percentage of Zero Friend Count = ', Percentage_of_friendships_Initiated_nill.round(decimals=2))

Percentage of Zero Friend Count =  3.02

• 3% of users were not initiated even a single friends

Analysis of Relation between Variables

Comparing Gender wise users in Age Group

fig,ax =plt.subplots(figsize=(10,8))
sns.set(font_scale=1)
sns.countplot(data = fb_df,x = 'age_group', hue='gender')
plt.title('Age vs Gender')

Text(0.5, 1.0, 'Age vs Gender')

• across all age group male users are more, except age group 51 to 80 where female users are more.

Gender vs Age

fig = sns.FacetGrid(fb_df,hue='gender',aspect=5)
fig.map(sns.kdeplot,'age',shade=True)
oldest = fb_df['age'].max()
fig.set(xlim=(0,oldest))
fig.add_legend()
plt.title('Age distribution using FacetGrid')

Text(0.5, 1.0, 'Age distribution using FacetGrid')

• Maximum Facebook Users of Males & Females are of Age 15 - 30 years.

Age vs Tenure

fig,ax =plt.subplots(figsize=(15,10))
sns.set(font_scale=1)
sns.boxplot(data=fb_df,x='age_group',y='tenure')

<AxesSubplot: xlabel='age_group', ylabel='tenure'>

sns.set(font_scale=1.5)
sns.set_palette(sns.color_palette("Set2", n_colors=5, desat=.5))
sns.catplot(x="age_group", y='tenure',hue ='gender',data=fb_df, kind="bar",height=8, aspect=2)

<seaborn.axisgrid.FacetGrid at 0x1e32d26ea40>

Max_tenure_in_year = (fb_df['tenure'].max())/365
Max_tenure_in_year

8.6

• Facebook launched on 4th of February 2004 since then users are using, maximum tenure as per the dataset is around 9 years, from this we can confirm this dataset is created in 2013.
• across all age group Female tenure is slightly higher than the Males.

Comparing Age vs Friend Count

sns.set(font_scale=1.5)
sns.set_palette(sns.color_palette("Set2", n_colors=5, desat=.5))
sns.catplot(x="age_group", y='friend_count',hue ='gender',data=fb_df, kind="bar",height=8, aspect=2)

<seaborn.axisgrid.FacetGrid at 0x1e32c2697e0>

• People with an age of less than 30 and more than 80 have more friends than the middle aged people.
• Females with an age of less than 30 have more friends than the Males.

How Friendships Initiated is Related to Age and Gender

fig,ax =plt.subplots(figsize=(10,5))
sns.set(font_scale=1)
sns.set_palette(sns.color_palette("Set2", n_colors=5, desat=.5))
sns.boxplot(data=fb_df,x='age_group',y='friendships_initiated')

<AxesSubplot: xlabel='age_group', ylabel='friendships_initiated'>

sns.set(font_scale=1.5)
sns.set_palette(sns.color_palette("Set2", n_colors=5, desat=.5))
sns.catplot(x="age_group", y='friendships_initiated',hue ='gender',data=fb_df, kind="bar",height=8, aspect=2)

<seaborn.axisgrid.FacetGrid at 0x1e32b2fb8e0>

• Upto 60 years of age Female Users were initiated more friends.
• Above 60 year of age Male Users were initiated more friends.

How does gender and age group related to Likes

Likes = ['likes', 'mobile_likes', 'www_likes', 'likes_received','mobile_likes_received','www_likes_received']
for value in Likes:
    sns.set(font_scale=1.5)
    sns.set_palette(sns.color_palette("Set2", n_colors=5, desat=.5))
    sns.catplot(x="age_group", y=value,hue ='gender',data=fb_df, kind="bar",height=8, aspect=2)

• Across the age group more Females are liked the Posts than the Males
• Age group 13 to 20 are more active in liking the Post
• The above pattern is reflecting the same in Likes Received also

Users through Website and Mobile App

Percentage_of_likes = (fb_df.likes.count() - fb_df['likes'].value_counts().max()) / (fb_df.likes.count())*100
Percentage_of_www_likes = (fb_df.likes.count() - fb_df['www_likes'].value_counts().max())/(fb_df.likes.count())*100
Percentage_of_mobile_likes = (fb_df.likes.count() -  fb_df['mobile_likes'].value_counts().max())/(fb_df.likes.count())*100
Percentage_of_likes_received = (fb_df.likes.count() - fb_df['likes_received'].value_counts().max())/(fb_df.likes_received.count())*100
Percentage_of_www_likes_received = (fb_df.likes.count() - fb_df['www_likes_received'].value_counts().max())/(fb_df.likes_received.count())*100
Percentage_of_mobile_likes_received = (fb_df.likes.count() - fb_df['mobile_likes_received'].value_counts().max())/(fb_df.likes_received.count())*100

print('Percentage of Likes = ',Percentage_of_likes)
print('Percentage of Mobile Likes = ',Percentage_of_mobile_likes)
print('Percentage of Website Likes = ',Percentage_of_www_likes)

print('Percentage of Likes Received = ',Percentage_of_likes_received)
print('Percentage of Mobile Likes Received = ',Percentage_of_mobile_likes_received)
print('Percentage of website Likes Received = ',Percentage_of_www_likes_received)

Percentage of Likes =  77.47360977827164
Percentage of Mobile Likes =  64.5961917268549
Percentage of Website Likes =  38.389817667558965
Percentage of Likes Received =  75.33208747916561
Percentage of Mobile Likes Received =  69.70048992373353
Percentage of website Likes Received =  62.76983686044749

Not_mobile_likes = 100- Percentage_of_mobile_likes
Not_www_likes = 100 - Percentage_of_www_likes
Not_mobile_likes_received = 100- Percentage_of_mobile_likes_received
Not_www_likes_received = 100 - Percentage_of_www_likes_received

labels = 'Yes','No'
size1 = [Percentage_of_mobile_likes,Not_mobile_likes]
size2 = [Percentage_of_www_likes,Not_www_likes]
size3 = [Percentage_of_mobile_likes_received,Not_mobile_likes_received]
size4 = [Percentage_of_www_likes_received,Not_www_likes_received]

colors = ['lightBlue', 'lightcoral']
plt.figure(figsize=(18,10), dpi=1600)

ax1 = plt.subplot2grid((2,2),(0,0))
plt.pie(size1,labels=labels, colors=colors,autopct='%1.1f%%',startangle=90)
plt.title('Users liked the Posts through Mobile App')

ax1 = plt.subplot2grid((2,2),(0,1))
plt.pie(size2,labels=labels, colors=colors,autopct='%1.1f%%',startangle=90)
plt.title('Users liked the Posts through Website')

ax1 = plt.subplot2grid((2,2),(1,0))
plt.pie(size3,labels=labels, colors=colors,autopct='%1.1f%%',startangle=90)
plt.title('Users Received the likes through Mobile App')

ax1 = plt.subplot2grid((2,2),(1,1))
plt.pie(size4,labels=labels, colors=colors,autopct='%1.1f%%',startangle=90)
plt.title('Users Received the likes through Website')

plt.axis('equal')
plt.show()

• 65% Users liked the Posts through Mobile App
• 70% Users liked the Posts through Website
• 38% Users Received the likes through Mobile App
• 63% Users Received the likes through Website

Friend Count and Friendships Initiated are Correlated

figsize=(15,10)
sns.scatterplot(x="friend_count", y="friendships_initiated", hue="gender", data = fb_df)

<AxesSubplot: xlabel='friend_count', ylabel='friendships_initiated'>

Friend Count and Tenure are Correlated

figsize=(15,10)
sns.scatterplot(x="friend_count", y="tenure", hue="gender", data = fb_df)

<AxesSubplot: xlabel='friend_count', ylabel='tenure'>

fb_df.corr() #CorreLation matrix

	age	dob_day	dob_year	dob_month	tenure	friend_count	friendships_initiated	likes	likes_received	mobile_likes	mobile_likes_received	www_likes	www_likes_received
age	1.000000	0.034978	-1.000000	0.025109	0.462685	-0.027429	-0.058084	-0.013020	-0.022575	-0.026725	-0.024253	0.015578	-0.018228
dob_day	0.034978	1.000000	-0.034978	0.129285	0.041759	0.021902	0.022932	0.015948	0.001355	0.014513	0.000485	0.009332	0.002449
dob_year	-1.000000	-0.034978	1.000000	-0.025109	-0.462685	0.027429	0.058084	0.013020	0.022575	0.026725	0.024253	-0.015578	0.018228
dob_month	0.025109	0.129285	-0.025109	1.000000	0.029343	0.019745	0.020009	0.014115	0.006484	0.010372	0.006424	0.012116	0.005992
tenure	0.462685	0.041759	-0.462685	0.029343	1.000000	0.166230	0.133474	0.057116	0.027739	0.028037	0.023965	0.070747	0.030547
friend_count	-0.027429	0.021902	0.027429	0.019745	0.166230	1.000000	0.825846	0.298010	0.236461	0.235650	0.232699	0.229798	0.220726
friendships_initiated	-0.058084	0.022932	0.058084	0.020009	0.133474	0.825846	1.000000	0.285584	0.175129	0.229800	0.173801	0.214017	0.161437
likes	-0.013020	0.015948	0.013020	0.014115	0.057116	0.298010	0.285584	1.000000	0.327375	0.871651	0.329257	0.644959	0.295686
likes_received	-0.022575	0.001355	0.022575	0.006484	0.027739	0.236461	0.175129	0.327375	1.000000	0.256996	0.973679	0.255364	0.947990
mobile_likes	-0.026725	0.014513	0.026725	0.010372	0.028037	0.235650	0.229800	0.871651	0.256996	1.000000	0.288512	0.187616	0.190171
mobile_likes_received	-0.024253	0.000485	0.024253	0.006424	0.023965	0.232699	0.173801	0.329257	0.973679	0.288512	1.000000	0.209996	0.850490
www_likes	0.015578	0.009332	-0.015578	0.012116	0.070747	0.229798	0.214017	0.644959	0.255364	0.187616	0.209996	1.000000	0.296052
www_likes_received	-0.018228	0.002449	0.018228	0.005992	0.030547	0.220726	0.161437	0.295686	0.947990	0.190171	0.850490	0.296052	1.000000

#PLotting Heat Map for the correlation Matrix.
correlations = fb_df.corr()
f, ax = plt.subplots(figsize=(22, 10))
sns.heatmap(data=correlations, annot = True, cmap='viridis')
sns.despine(left=True, bottom=True)

Observations :

• Friend Count and Friendships Initiated have strong Correlations, as Friendships Initiated increases and the Friend Count tends to increase.

Initial glance it is observed that there are two & three dark colored squares that get my attention which is 'likes' & 'mobile_likes', 'likes' & 'www_likes', 'likes_received' & 'mobile_likes_received', 'likes_received' & 'www_likes_received', 'www_likes_received' & 'mobile_likes_received', ('tenure', 'age' & 'dob_year') and 'friend_count' & 'friendships_initiated'

'likes' & 'mobile_likes' and 'likes' & 'www_likes' are also closely correlated so one of them we can keep for futher analysis and other 2 features can be dropped 'mobile_likes' and 'www_likes'.

'likes' is strongly correlated with 'mobile_likes' and 'www_likes' as sum of www_likes and mobile_likes is 'likes' column. Similarly 'likes_received' is strongly correlated with 'mobile_likes_received' and 'www_likes_received' as sum of www_likes_received and mobile_likes_received is 'likes' column.

'friend_count' & 'friendship_initiated' are closely correlated hence 'friendship_initiated' column is excluded. 'dob_year' is negatively correlated with 'age' and 'tenure' also 'age' & 'tenure' strongly correlated. 'dob_month' and 'dob_day' are not depend on other features hence we are excluding 'dob_year', 'dob_month' and 'dob_day'.

df = fb_df[['age','tenure', 'friend_count','likes','likes_received','gender']]
df.set_index('age', inplace=True)
corr = df.corr()

# Set up the matplotlib figure
f, ax = plt.subplots(figsize=(5, 4))

# Generate a mask for the upper triangle
mask = np.zeros_like(corr, dtype=np.bool)
mask[np.triu_indices_from(mask)] = True

# Generate a custom diverging colormap
cmap = sns.diverging_palette(220, 10, as_cmap=True)

# Draw the heatmap with the mask and correct aspect ratio
g= sns.heatmap(corr, mask=mask, cmap=cmap, vmax=.3, center=0,
            square=True, linewidths=.5, cbar_kws={"shrink": .5})

sns.clustermap(corr, center=0, cmap="vlag",linewidths=.75, figsize=(4, 5))

<seaborn.matrix.ClusterGrid at 0x1e3232d9210>

sns.pairplot(fb_df[['age','gender','tenure','friend_count','friendships_initiated','likes']],
             vars= ['age','tenure','friend_count','friendships_initiated','likes'],hue="gender", palette="husl")
plt.title('Pair Plot')

Text(0.5, 1.0, 'Pair Plot')

---

9. Summary and Conclusion

---

This section showcases the Summary and the conclusion from the exploration done on the dataset

• Maximum users of Facebook are of Age 15 to 30 years.
• Maximum Age:113 years and Minimum Age:13 years
• Maximum users of Facebook are Males, about 59.3%.
• Even though the male users are more in number but Female users are more active in Facebook.
• More number of User have joined from Last 2 Years of Data collection
• more number of users have fewer than 500 Facebook Friends
• Across all age group male users are more, except age group 51 to 80 where female users are more.
• People with an age of less than 30 and more than 80 have more friends than the middle aged people.
• Females with an age of less than 30 have more friends than the Males
• Facebook launched on 4th of February 2004 since then users are using, maximum tenure as per the dataset is around 9 years, from this we can confirm this dataset is created in 2013.
• across all age group male and female users tenure is almost same
• Across the age group more Females are liked the Posts than the Males
• Age group 13 to 20 are more active in liking the Post
• The above pattern is reflecting the same in Likes Received also
• 65% Users liked the Posts through Mobile App
• 70% Users liked the Posts through Website
• 38% Users Received the likes through Mobile App
• 63% Users Received the likes through Website
• Upto 60 years of age Female Users were initiated more friends.
• Above 60 year of age Male Users were initiated more friends.